1 #ifndef _LINUX_MM_H 2 #define _LINUX_MM_H 3 4 #include <linux/errno.h> 5 6 #ifdef __KERNEL__ 7 8 #include <linux/gfp.h> 9 #include <linux/list.h> 10 #include <linux/mmzone.h> 11 #include <linux/rbtree.h> 12 #include <linux/prio_tree.h> 13 #include <linux/debug_locks.h> 14 #include <linux/mm_types.h> 15 #include <linux/range.h> 16 17 struct mempolicy; 18 struct anon_vma; 19 struct file_ra_state; 20 struct user_struct; 21 struct writeback_control; 22 23 #ifndef CONFIG_DISCONTIGMEM /* Don't use mapnrs, do it properly */ 24 extern unsigned long max_mapnr; 25 #endif 26 27 extern unsigned long num_physpages; 28 extern unsigned long totalram_pages; 29 extern void * high_memory; 30 extern int page_cluster; 31 32 #ifdef CONFIG_SYSCTL 33 extern int sysctl_legacy_va_layout; 34 #else 35 #define sysctl_legacy_va_layout 0 36 #endif 37 38 #include <asm/page.h> 39 #include <asm/pgtable.h> 40 #include <asm/processor.h> 41 42 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n)) 43 44 /* to align the pointer to the (next) page boundary */ 45 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE) 46 47 /* 48 * Linux kernel virtual memory manager primitives. 49 * The idea being to have a "virtual" mm in the same way 50 * we have a virtual fs - giving a cleaner interface to the 51 * mm details, and allowing different kinds of memory mappings 52 * (from shared memory to executable loading to arbitrary 53 * mmap() functions). 54 */ 55 56 extern struct kmem_cache *vm_area_cachep; 57 58 #ifndef CONFIG_MMU 59 extern struct rb_root nommu_region_tree; 60 extern struct rw_semaphore nommu_region_sem; 61 62 extern unsigned int kobjsize(const void *objp); 63 #endif 64 65 /* 66 * vm_flags in vm_area_struct, see mm_types.h. 67 */ 68 #define VM_READ 0x00000001 /* currently active flags */ 69 #define VM_WRITE 0x00000002 70 #define VM_EXEC 0x00000004 71 #define VM_SHARED 0x00000008 72 73 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */ 74 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */ 75 #define VM_MAYWRITE 0x00000020 76 #define VM_MAYEXEC 0x00000040 77 #define VM_MAYSHARE 0x00000080 78 79 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */ 80 #define VM_GROWSUP 0x00000200 81 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */ 82 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */ 83 84 #define VM_EXECUTABLE 0x00001000 85 #define VM_LOCKED 0x00002000 86 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */ 87 88 /* Used by sys_madvise() */ 89 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */ 90 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */ 91 92 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */ 93 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */ 94 #define VM_RESERVED 0x00080000 /* Count as reserved_vm like IO */ 95 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */ 96 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */ 97 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */ 98 #define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */ 99 #define VM_MAPPED_COPY 0x01000000 /* T if mapped copy of data (nommu mmap) */ 100 #define VM_INSERTPAGE 0x02000000 /* The vma has had "vm_insert_page()" done on it */ 101 #define VM_ALWAYSDUMP 0x04000000 /* Always include in core dumps */ 102 103 #define VM_CAN_NONLINEAR 0x08000000 /* Has ->fault & does nonlinear pages */ 104 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */ 105 #define VM_SAO 0x20000000 /* Strong Access Ordering (powerpc) */ 106 #define VM_PFN_AT_MMAP 0x40000000 /* PFNMAP vma that is fully mapped at mmap time */ 107 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */ 108 109 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */ 110 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS 111 #endif 112 113 #ifdef CONFIG_STACK_GROWSUP 114 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT) 115 #else 116 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT) 117 #endif 118 119 #define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ) 120 #define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK 121 #define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK)) 122 #define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ) 123 #define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ) 124 125 /* 126 * special vmas that are non-mergable, non-mlock()able 127 */ 128 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_RESERVED | VM_PFNMAP) 129 130 /* 131 * mapping from the currently active vm_flags protection bits (the 132 * low four bits) to a page protection mask.. 133 */ 134 extern pgprot_t protection_map[16]; 135 136 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */ 137 #define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */ 138 #define FAULT_FLAG_MKWRITE 0x04 /* Fault was mkwrite of existing pte */ 139 140 /* 141 * This interface is used by x86 PAT code to identify a pfn mapping that is 142 * linear over entire vma. This is to optimize PAT code that deals with 143 * marking the physical region with a particular prot. This is not for generic 144 * mm use. Note also that this check will not work if the pfn mapping is 145 * linear for a vma starting at physical address 0. In which case PAT code 146 * falls back to slow path of reserving physical range page by page. 147 */ 148 static inline int is_linear_pfn_mapping(struct vm_area_struct *vma) 149 { 150 return (vma->vm_flags & VM_PFN_AT_MMAP); 151 } 152 153 static inline int is_pfn_mapping(struct vm_area_struct *vma) 154 { 155 return (vma->vm_flags & VM_PFNMAP); 156 } 157 158 /* 159 * vm_fault is filled by the the pagefault handler and passed to the vma's 160 * ->fault function. The vma's ->fault is responsible for returning a bitmask 161 * of VM_FAULT_xxx flags that give details about how the fault was handled. 162 * 163 * pgoff should be used in favour of virtual_address, if possible. If pgoff 164 * is used, one may set VM_CAN_NONLINEAR in the vma->vm_flags to get nonlinear 165 * mapping support. 166 */ 167 struct vm_fault { 168 unsigned int flags; /* FAULT_FLAG_xxx flags */ 169 pgoff_t pgoff; /* Logical page offset based on vma */ 170 void __user *virtual_address; /* Faulting virtual address */ 171 172 struct page *page; /* ->fault handlers should return a 173 * page here, unless VM_FAULT_NOPAGE 174 * is set (which is also implied by 175 * VM_FAULT_ERROR). 176 */ 177 }; 178 179 /* 180 * These are the virtual MM functions - opening of an area, closing and 181 * unmapping it (needed to keep files on disk up-to-date etc), pointer 182 * to the functions called when a no-page or a wp-page exception occurs. 183 */ 184 struct vm_operations_struct { 185 void (*open)(struct vm_area_struct * area); 186 void (*close)(struct vm_area_struct * area); 187 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf); 188 189 /* notification that a previously read-only page is about to become 190 * writable, if an error is returned it will cause a SIGBUS */ 191 int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf); 192 193 /* called by access_process_vm when get_user_pages() fails, typically 194 * for use by special VMAs that can switch between memory and hardware 195 */ 196 int (*access)(struct vm_area_struct *vma, unsigned long addr, 197 void *buf, int len, int write); 198 #ifdef CONFIG_NUMA 199 /* 200 * set_policy() op must add a reference to any non-NULL @new mempolicy 201 * to hold the policy upon return. Caller should pass NULL @new to 202 * remove a policy and fall back to surrounding context--i.e. do not 203 * install a MPOL_DEFAULT policy, nor the task or system default 204 * mempolicy. 205 */ 206 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new); 207 208 /* 209 * get_policy() op must add reference [mpol_get()] to any policy at 210 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure 211 * in mm/mempolicy.c will do this automatically. 212 * get_policy() must NOT add a ref if the policy at (vma,addr) is not 213 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem. 214 * If no [shared/vma] mempolicy exists at the addr, get_policy() op 215 * must return NULL--i.e., do not "fallback" to task or system default 216 * policy. 217 */ 218 struct mempolicy *(*get_policy)(struct vm_area_struct *vma, 219 unsigned long addr); 220 int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from, 221 const nodemask_t *to, unsigned long flags); 222 #endif 223 }; 224 225 struct mmu_gather; 226 struct inode; 227 228 #define page_private(page) ((page)->private) 229 #define set_page_private(page, v) ((page)->private = (v)) 230 231 /* 232 * FIXME: take this include out, include page-flags.h in 233 * files which need it (119 of them) 234 */ 235 #include <linux/page-flags.h> 236 237 /* 238 * Methods to modify the page usage count. 239 * 240 * What counts for a page usage: 241 * - cache mapping (page->mapping) 242 * - private data (page->private) 243 * - page mapped in a task's page tables, each mapping 244 * is counted separately 245 * 246 * Also, many kernel routines increase the page count before a critical 247 * routine so they can be sure the page doesn't go away from under them. 248 */ 249 250 /* 251 * Drop a ref, return true if the refcount fell to zero (the page has no users) 252 */ 253 static inline int put_page_testzero(struct page *page) 254 { 255 VM_BUG_ON(atomic_read(&page->_count) == 0); 256 return atomic_dec_and_test(&page->_count); 257 } 258 259 /* 260 * Try to grab a ref unless the page has a refcount of zero, return false if 261 * that is the case. 262 */ 263 static inline int get_page_unless_zero(struct page *page) 264 { 265 return atomic_inc_not_zero(&page->_count); 266 } 267 268 extern int page_is_ram(unsigned long pfn); 269 270 /* Support for virtually mapped pages */ 271 struct page *vmalloc_to_page(const void *addr); 272 unsigned long vmalloc_to_pfn(const void *addr); 273 274 /* 275 * Determine if an address is within the vmalloc range 276 * 277 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there 278 * is no special casing required. 279 */ 280 static inline int is_vmalloc_addr(const void *x) 281 { 282 #ifdef CONFIG_MMU 283 unsigned long addr = (unsigned long)x; 284 285 return addr >= VMALLOC_START && addr < VMALLOC_END; 286 #else 287 return 0; 288 #endif 289 } 290 #ifdef CONFIG_MMU 291 extern int is_vmalloc_or_module_addr(const void *x); 292 #else 293 static inline int is_vmalloc_or_module_addr(const void *x) 294 { 295 return 0; 296 } 297 #endif 298 299 static inline struct page *compound_head(struct page *page) 300 { 301 if (unlikely(PageTail(page))) 302 return page->first_page; 303 return page; 304 } 305 306 static inline int page_count(struct page *page) 307 { 308 return atomic_read(&compound_head(page)->_count); 309 } 310 311 static inline void get_page(struct page *page) 312 { 313 page = compound_head(page); 314 VM_BUG_ON(atomic_read(&page->_count) == 0); 315 atomic_inc(&page->_count); 316 } 317 318 static inline struct page *virt_to_head_page(const void *x) 319 { 320 struct page *page = virt_to_page(x); 321 return compound_head(page); 322 } 323 324 /* 325 * Setup the page count before being freed into the page allocator for 326 * the first time (boot or memory hotplug) 327 */ 328 static inline void init_page_count(struct page *page) 329 { 330 atomic_set(&page->_count, 1); 331 } 332 333 void put_page(struct page *page); 334 void put_pages_list(struct list_head *pages); 335 336 void split_page(struct page *page, unsigned int order); 337 338 /* 339 * Compound pages have a destructor function. Provide a 340 * prototype for that function and accessor functions. 341 * These are _only_ valid on the head of a PG_compound page. 342 */ 343 typedef void compound_page_dtor(struct page *); 344 345 static inline void set_compound_page_dtor(struct page *page, 346 compound_page_dtor *dtor) 347 { 348 page[1].lru.next = (void *)dtor; 349 } 350 351 static inline compound_page_dtor *get_compound_page_dtor(struct page *page) 352 { 353 return (compound_page_dtor *)page[1].lru.next; 354 } 355 356 static inline int compound_order(struct page *page) 357 { 358 if (!PageHead(page)) 359 return 0; 360 return (unsigned long)page[1].lru.prev; 361 } 362 363 static inline void set_compound_order(struct page *page, unsigned long order) 364 { 365 page[1].lru.prev = (void *)order; 366 } 367 368 /* 369 * Multiple processes may "see" the same page. E.g. for untouched 370 * mappings of /dev/null, all processes see the same page full of 371 * zeroes, and text pages of executables and shared libraries have 372 * only one copy in memory, at most, normally. 373 * 374 * For the non-reserved pages, page_count(page) denotes a reference count. 375 * page_count() == 0 means the page is free. page->lru is then used for 376 * freelist management in the buddy allocator. 377 * page_count() > 0 means the page has been allocated. 378 * 379 * Pages are allocated by the slab allocator in order to provide memory 380 * to kmalloc and kmem_cache_alloc. In this case, the management of the 381 * page, and the fields in 'struct page' are the responsibility of mm/slab.c 382 * unless a particular usage is carefully commented. (the responsibility of 383 * freeing the kmalloc memory is the caller's, of course). 384 * 385 * A page may be used by anyone else who does a __get_free_page(). 386 * In this case, page_count still tracks the references, and should only 387 * be used through the normal accessor functions. The top bits of page->flags 388 * and page->virtual store page management information, but all other fields 389 * are unused and could be used privately, carefully. The management of this 390 * page is the responsibility of the one who allocated it, and those who have 391 * subsequently been given references to it. 392 * 393 * The other pages (we may call them "pagecache pages") are completely 394 * managed by the Linux memory manager: I/O, buffers, swapping etc. 395 * The following discussion applies only to them. 396 * 397 * A pagecache page contains an opaque `private' member, which belongs to the 398 * page's address_space. Usually, this is the address of a circular list of 399 * the page's disk buffers. PG_private must be set to tell the VM to call 400 * into the filesystem to release these pages. 401 * 402 * A page may belong to an inode's memory mapping. In this case, page->mapping 403 * is the pointer to the inode, and page->index is the file offset of the page, 404 * in units of PAGE_CACHE_SIZE. 405 * 406 * If pagecache pages are not associated with an inode, they are said to be 407 * anonymous pages. These may become associated with the swapcache, and in that 408 * case PG_swapcache is set, and page->private is an offset into the swapcache. 409 * 410 * In either case (swapcache or inode backed), the pagecache itself holds one 411 * reference to the page. Setting PG_private should also increment the 412 * refcount. The each user mapping also has a reference to the page. 413 * 414 * The pagecache pages are stored in a per-mapping radix tree, which is 415 * rooted at mapping->page_tree, and indexed by offset. 416 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space 417 * lists, we instead now tag pages as dirty/writeback in the radix tree. 418 * 419 * All pagecache pages may be subject to I/O: 420 * - inode pages may need to be read from disk, 421 * - inode pages which have been modified and are MAP_SHARED may need 422 * to be written back to the inode on disk, 423 * - anonymous pages (including MAP_PRIVATE file mappings) which have been 424 * modified may need to be swapped out to swap space and (later) to be read 425 * back into memory. 426 */ 427 428 /* 429 * The zone field is never updated after free_area_init_core() 430 * sets it, so none of the operations on it need to be atomic. 431 */ 432 433 434 /* 435 * page->flags layout: 436 * 437 * There are three possibilities for how page->flags get 438 * laid out. The first is for the normal case, without 439 * sparsemem. The second is for sparsemem when there is 440 * plenty of space for node and section. The last is when 441 * we have run out of space and have to fall back to an 442 * alternate (slower) way of determining the node. 443 * 444 * No sparsemem or sparsemem vmemmap: | NODE | ZONE | ... | FLAGS | 445 * classic sparse with space for node:| SECTION | NODE | ZONE | ... | FLAGS | 446 * classic sparse no space for node: | SECTION | ZONE | ... | FLAGS | 447 */ 448 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP) 449 #define SECTIONS_WIDTH SECTIONS_SHIFT 450 #else 451 #define SECTIONS_WIDTH 0 452 #endif 453 454 #define ZONES_WIDTH ZONES_SHIFT 455 456 #if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= BITS_PER_LONG - NR_PAGEFLAGS 457 #define NODES_WIDTH NODES_SHIFT 458 #else 459 #ifdef CONFIG_SPARSEMEM_VMEMMAP 460 #error "Vmemmap: No space for nodes field in page flags" 461 #endif 462 #define NODES_WIDTH 0 463 #endif 464 465 /* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */ 466 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH) 467 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH) 468 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH) 469 470 /* 471 * We are going to use the flags for the page to node mapping if its in 472 * there. This includes the case where there is no node, so it is implicit. 473 */ 474 #if !(NODES_WIDTH > 0 || NODES_SHIFT == 0) 475 #define NODE_NOT_IN_PAGE_FLAGS 476 #endif 477 478 #ifndef PFN_SECTION_SHIFT 479 #define PFN_SECTION_SHIFT 0 480 #endif 481 482 /* 483 * Define the bit shifts to access each section. For non-existant 484 * sections we define the shift as 0; that plus a 0 mask ensures 485 * the compiler will optimise away reference to them. 486 */ 487 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0)) 488 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0)) 489 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0)) 490 491 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allcator */ 492 #ifdef NODE_NOT_IN_PAGEFLAGS 493 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT) 494 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \ 495 SECTIONS_PGOFF : ZONES_PGOFF) 496 #else 497 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT) 498 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \ 499 NODES_PGOFF : ZONES_PGOFF) 500 #endif 501 502 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0)) 503 504 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS 505 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS 506 #endif 507 508 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1) 509 #define NODES_MASK ((1UL << NODES_WIDTH) - 1) 510 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1) 511 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1) 512 513 static inline enum zone_type page_zonenum(struct page *page) 514 { 515 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK; 516 } 517 518 /* 519 * The identification function is only used by the buddy allocator for 520 * determining if two pages could be buddies. We are not really 521 * identifying a zone since we could be using a the section number 522 * id if we have not node id available in page flags. 523 * We guarantee only that it will return the same value for two 524 * combinable pages in a zone. 525 */ 526 static inline int page_zone_id(struct page *page) 527 { 528 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK; 529 } 530 531 static inline int zone_to_nid(struct zone *zone) 532 { 533 #ifdef CONFIG_NUMA 534 return zone->node; 535 #else 536 return 0; 537 #endif 538 } 539 540 #ifdef NODE_NOT_IN_PAGE_FLAGS 541 extern int page_to_nid(struct page *page); 542 #else 543 static inline int page_to_nid(struct page *page) 544 { 545 return (page->flags >> NODES_PGSHIFT) & NODES_MASK; 546 } 547 #endif 548 549 static inline struct zone *page_zone(struct page *page) 550 { 551 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)]; 552 } 553 554 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP) 555 static inline unsigned long page_to_section(struct page *page) 556 { 557 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK; 558 } 559 #endif 560 561 static inline void set_page_zone(struct page *page, enum zone_type zone) 562 { 563 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT); 564 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT; 565 } 566 567 static inline void set_page_node(struct page *page, unsigned long node) 568 { 569 page->flags &= ~(NODES_MASK << NODES_PGSHIFT); 570 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT; 571 } 572 573 static inline void set_page_section(struct page *page, unsigned long section) 574 { 575 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT); 576 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT; 577 } 578 579 static inline void set_page_links(struct page *page, enum zone_type zone, 580 unsigned long node, unsigned long pfn) 581 { 582 set_page_zone(page, zone); 583 set_page_node(page, node); 584 set_page_section(page, pfn_to_section_nr(pfn)); 585 } 586 587 /* 588 * Some inline functions in vmstat.h depend on page_zone() 589 */ 590 #include <linux/vmstat.h> 591 592 static __always_inline void *lowmem_page_address(struct page *page) 593 { 594 return __va(page_to_pfn(page) << PAGE_SHIFT); 595 } 596 597 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL) 598 #define HASHED_PAGE_VIRTUAL 599 #endif 600 601 #if defined(WANT_PAGE_VIRTUAL) 602 #define page_address(page) ((page)->virtual) 603 #define set_page_address(page, address) \ 604 do { \ 605 (page)->virtual = (address); \ 606 } while(0) 607 #define page_address_init() do { } while(0) 608 #endif 609 610 #if defined(HASHED_PAGE_VIRTUAL) 611 void *page_address(struct page *page); 612 void set_page_address(struct page *page, void *virtual); 613 void page_address_init(void); 614 #endif 615 616 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL) 617 #define page_address(page) lowmem_page_address(page) 618 #define set_page_address(page, address) do { } while(0) 619 #define page_address_init() do { } while(0) 620 #endif 621 622 /* 623 * On an anonymous page mapped into a user virtual memory area, 624 * page->mapping points to its anon_vma, not to a struct address_space; 625 * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h. 626 * 627 * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled, 628 * the PAGE_MAPPING_KSM bit may be set along with the PAGE_MAPPING_ANON bit; 629 * and then page->mapping points, not to an anon_vma, but to a private 630 * structure which KSM associates with that merged page. See ksm.h. 631 * 632 * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is currently never used. 633 * 634 * Please note that, confusingly, "page_mapping" refers to the inode 635 * address_space which maps the page from disk; whereas "page_mapped" 636 * refers to user virtual address space into which the page is mapped. 637 */ 638 #define PAGE_MAPPING_ANON 1 639 #define PAGE_MAPPING_KSM 2 640 #define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_KSM) 641 642 extern struct address_space swapper_space; 643 static inline struct address_space *page_mapping(struct page *page) 644 { 645 struct address_space *mapping = page->mapping; 646 647 VM_BUG_ON(PageSlab(page)); 648 if (unlikely(PageSwapCache(page))) 649 mapping = &swapper_space; 650 else if (unlikely((unsigned long)mapping & PAGE_MAPPING_ANON)) 651 mapping = NULL; 652 return mapping; 653 } 654 655 /* Neutral page->mapping pointer to address_space or anon_vma or other */ 656 static inline void *page_rmapping(struct page *page) 657 { 658 return (void *)((unsigned long)page->mapping & ~PAGE_MAPPING_FLAGS); 659 } 660 661 static inline int PageAnon(struct page *page) 662 { 663 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0; 664 } 665 666 /* 667 * Return the pagecache index of the passed page. Regular pagecache pages 668 * use ->index whereas swapcache pages use ->private 669 */ 670 static inline pgoff_t page_index(struct page *page) 671 { 672 if (unlikely(PageSwapCache(page))) 673 return page_private(page); 674 return page->index; 675 } 676 677 /* 678 * The atomic page->_mapcount, like _count, starts from -1: 679 * so that transitions both from it and to it can be tracked, 680 * using atomic_inc_and_test and atomic_add_negative(-1). 681 */ 682 static inline void reset_page_mapcount(struct page *page) 683 { 684 atomic_set(&(page)->_mapcount, -1); 685 } 686 687 static inline int page_mapcount(struct page *page) 688 { 689 return atomic_read(&(page)->_mapcount) + 1; 690 } 691 692 /* 693 * Return true if this page is mapped into pagetables. 694 */ 695 static inline int page_mapped(struct page *page) 696 { 697 return atomic_read(&(page)->_mapcount) >= 0; 698 } 699 700 /* 701 * Different kinds of faults, as returned by handle_mm_fault(). 702 * Used to decide whether a process gets delivered SIGBUS or 703 * just gets major/minor fault counters bumped up. 704 */ 705 706 #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */ 707 708 #define VM_FAULT_OOM 0x0001 709 #define VM_FAULT_SIGBUS 0x0002 710 #define VM_FAULT_MAJOR 0x0004 711 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */ 712 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned page */ 713 714 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */ 715 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */ 716 717 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_HWPOISON) 718 719 /* 720 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM. 721 */ 722 extern void pagefault_out_of_memory(void); 723 724 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK) 725 726 extern void show_free_areas(void); 727 728 int shmem_lock(struct file *file, int lock, struct user_struct *user); 729 struct file *shmem_file_setup(const char *name, loff_t size, unsigned long flags); 730 int shmem_zero_setup(struct vm_area_struct *); 731 732 #ifndef CONFIG_MMU 733 extern unsigned long shmem_get_unmapped_area(struct file *file, 734 unsigned long addr, 735 unsigned long len, 736 unsigned long pgoff, 737 unsigned long flags); 738 #endif 739 740 extern int can_do_mlock(void); 741 extern int user_shm_lock(size_t, struct user_struct *); 742 extern void user_shm_unlock(size_t, struct user_struct *); 743 744 /* 745 * Parameter block passed down to zap_pte_range in exceptional cases. 746 */ 747 struct zap_details { 748 struct vm_area_struct *nonlinear_vma; /* Check page->index if set */ 749 struct address_space *check_mapping; /* Check page->mapping if set */ 750 pgoff_t first_index; /* Lowest page->index to unmap */ 751 pgoff_t last_index; /* Highest page->index to unmap */ 752 spinlock_t *i_mmap_lock; /* For unmap_mapping_range: */ 753 unsigned long truncate_count; /* Compare vm_truncate_count */ 754 }; 755 756 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr, 757 pte_t pte); 758 759 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address, 760 unsigned long size); 761 unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address, 762 unsigned long size, struct zap_details *); 763 unsigned long unmap_vmas(struct mmu_gather **tlb, 764 struct vm_area_struct *start_vma, unsigned long start_addr, 765 unsigned long end_addr, unsigned long *nr_accounted, 766 struct zap_details *); 767 768 /** 769 * mm_walk - callbacks for walk_page_range 770 * @pgd_entry: if set, called for each non-empty PGD (top-level) entry 771 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry 772 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry 773 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry 774 * @pte_hole: if set, called for each hole at all levels 775 * @hugetlb_entry: if set, called for each hugetlb entry 776 * 777 * (see walk_page_range for more details) 778 */ 779 struct mm_walk { 780 int (*pgd_entry)(pgd_t *, unsigned long, unsigned long, struct mm_walk *); 781 int (*pud_entry)(pud_t *, unsigned long, unsigned long, struct mm_walk *); 782 int (*pmd_entry)(pmd_t *, unsigned long, unsigned long, struct mm_walk *); 783 int (*pte_entry)(pte_t *, unsigned long, unsigned long, struct mm_walk *); 784 int (*pte_hole)(unsigned long, unsigned long, struct mm_walk *); 785 int (*hugetlb_entry)(pte_t *, unsigned long, 786 unsigned long, unsigned long, struct mm_walk *); 787 struct mm_struct *mm; 788 void *private; 789 }; 790 791 int walk_page_range(unsigned long addr, unsigned long end, 792 struct mm_walk *walk); 793 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr, 794 unsigned long end, unsigned long floor, unsigned long ceiling); 795 int copy_page_range(struct mm_struct *dst, struct mm_struct *src, 796 struct vm_area_struct *vma); 797 void unmap_mapping_range(struct address_space *mapping, 798 loff_t const holebegin, loff_t const holelen, int even_cows); 799 int follow_pfn(struct vm_area_struct *vma, unsigned long address, 800 unsigned long *pfn); 801 int follow_phys(struct vm_area_struct *vma, unsigned long address, 802 unsigned int flags, unsigned long *prot, resource_size_t *phys); 803 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr, 804 void *buf, int len, int write); 805 806 static inline void unmap_shared_mapping_range(struct address_space *mapping, 807 loff_t const holebegin, loff_t const holelen) 808 { 809 unmap_mapping_range(mapping, holebegin, holelen, 0); 810 } 811 812 extern void truncate_pagecache(struct inode *inode, loff_t old, loff_t new); 813 extern int vmtruncate(struct inode *inode, loff_t offset); 814 extern int vmtruncate_range(struct inode *inode, loff_t offset, loff_t end); 815 816 int truncate_inode_page(struct address_space *mapping, struct page *page); 817 int generic_error_remove_page(struct address_space *mapping, struct page *page); 818 819 int invalidate_inode_page(struct page *page); 820 821 #ifdef CONFIG_MMU 822 extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma, 823 unsigned long address, unsigned int flags); 824 #else 825 static inline int handle_mm_fault(struct mm_struct *mm, 826 struct vm_area_struct *vma, unsigned long address, 827 unsigned int flags) 828 { 829 /* should never happen if there's no MMU */ 830 BUG(); 831 return VM_FAULT_SIGBUS; 832 } 833 #endif 834 835 extern int make_pages_present(unsigned long addr, unsigned long end); 836 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write); 837 838 int get_user_pages(struct task_struct *tsk, struct mm_struct *mm, 839 unsigned long start, int nr_pages, int write, int force, 840 struct page **pages, struct vm_area_struct **vmas); 841 int get_user_pages_fast(unsigned long start, int nr_pages, int write, 842 struct page **pages); 843 struct page *get_dump_page(unsigned long addr); 844 845 extern int try_to_release_page(struct page * page, gfp_t gfp_mask); 846 extern void do_invalidatepage(struct page *page, unsigned long offset); 847 848 int __set_page_dirty_nobuffers(struct page *page); 849 int __set_page_dirty_no_writeback(struct page *page); 850 int redirty_page_for_writepage(struct writeback_control *wbc, 851 struct page *page); 852 void account_page_dirtied(struct page *page, struct address_space *mapping); 853 int set_page_dirty(struct page *page); 854 int set_page_dirty_lock(struct page *page); 855 int clear_page_dirty_for_io(struct page *page); 856 857 extern unsigned long move_page_tables(struct vm_area_struct *vma, 858 unsigned long old_addr, struct vm_area_struct *new_vma, 859 unsigned long new_addr, unsigned long len); 860 extern unsigned long do_mremap(unsigned long addr, 861 unsigned long old_len, unsigned long new_len, 862 unsigned long flags, unsigned long new_addr); 863 extern int mprotect_fixup(struct vm_area_struct *vma, 864 struct vm_area_struct **pprev, unsigned long start, 865 unsigned long end, unsigned long newflags); 866 867 /* 868 * doesn't attempt to fault and will return short. 869 */ 870 int __get_user_pages_fast(unsigned long start, int nr_pages, int write, 871 struct page **pages); 872 /* 873 * per-process(per-mm_struct) statistics. 874 */ 875 #if defined(SPLIT_RSS_COUNTING) 876 /* 877 * The mm counters are not protected by its page_table_lock, 878 * so must be incremented atomically. 879 */ 880 static inline void set_mm_counter(struct mm_struct *mm, int member, long value) 881 { 882 atomic_long_set(&mm->rss_stat.count[member], value); 883 } 884 885 unsigned long get_mm_counter(struct mm_struct *mm, int member); 886 887 static inline void add_mm_counter(struct mm_struct *mm, int member, long value) 888 { 889 atomic_long_add(value, &mm->rss_stat.count[member]); 890 } 891 892 static inline void inc_mm_counter(struct mm_struct *mm, int member) 893 { 894 atomic_long_inc(&mm->rss_stat.count[member]); 895 } 896 897 static inline void dec_mm_counter(struct mm_struct *mm, int member) 898 { 899 atomic_long_dec(&mm->rss_stat.count[member]); 900 } 901 902 #else /* !USE_SPLIT_PTLOCKS */ 903 /* 904 * The mm counters are protected by its page_table_lock, 905 * so can be incremented directly. 906 */ 907 static inline void set_mm_counter(struct mm_struct *mm, int member, long value) 908 { 909 mm->rss_stat.count[member] = value; 910 } 911 912 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member) 913 { 914 return mm->rss_stat.count[member]; 915 } 916 917 static inline void add_mm_counter(struct mm_struct *mm, int member, long value) 918 { 919 mm->rss_stat.count[member] += value; 920 } 921 922 static inline void inc_mm_counter(struct mm_struct *mm, int member) 923 { 924 mm->rss_stat.count[member]++; 925 } 926 927 static inline void dec_mm_counter(struct mm_struct *mm, int member) 928 { 929 mm->rss_stat.count[member]--; 930 } 931 932 #endif /* !USE_SPLIT_PTLOCKS */ 933 934 static inline unsigned long get_mm_rss(struct mm_struct *mm) 935 { 936 return get_mm_counter(mm, MM_FILEPAGES) + 937 get_mm_counter(mm, MM_ANONPAGES); 938 } 939 940 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm) 941 { 942 return max(mm->hiwater_rss, get_mm_rss(mm)); 943 } 944 945 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm) 946 { 947 return max(mm->hiwater_vm, mm->total_vm); 948 } 949 950 static inline void update_hiwater_rss(struct mm_struct *mm) 951 { 952 unsigned long _rss = get_mm_rss(mm); 953 954 if ((mm)->hiwater_rss < _rss) 955 (mm)->hiwater_rss = _rss; 956 } 957 958 static inline void update_hiwater_vm(struct mm_struct *mm) 959 { 960 if (mm->hiwater_vm < mm->total_vm) 961 mm->hiwater_vm = mm->total_vm; 962 } 963 964 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss, 965 struct mm_struct *mm) 966 { 967 unsigned long hiwater_rss = get_mm_hiwater_rss(mm); 968 969 if (*maxrss < hiwater_rss) 970 *maxrss = hiwater_rss; 971 } 972 973 #if defined(SPLIT_RSS_COUNTING) 974 void sync_mm_rss(struct task_struct *task, struct mm_struct *mm); 975 #else 976 static inline void sync_mm_rss(struct task_struct *task, struct mm_struct *mm) 977 { 978 } 979 #endif 980 981 /* 982 * A callback you can register to apply pressure to ageable caches. 983 * 984 * 'shrink' is passed a count 'nr_to_scan' and a 'gfpmask'. It should 985 * look through the least-recently-used 'nr_to_scan' entries and 986 * attempt to free them up. It should return the number of objects 987 * which remain in the cache. If it returns -1, it means it cannot do 988 * any scanning at this time (eg. there is a risk of deadlock). 989 * 990 * The 'gfpmask' refers to the allocation we are currently trying to 991 * fulfil. 992 * 993 * Note that 'shrink' will be passed nr_to_scan == 0 when the VM is 994 * querying the cache size, so a fastpath for that case is appropriate. 995 */ 996 struct shrinker { 997 int (*shrink)(int nr_to_scan, gfp_t gfp_mask); 998 int seeks; /* seeks to recreate an obj */ 999 1000 /* These are for internal use */ 1001 struct list_head list; 1002 long nr; /* objs pending delete */ 1003 }; 1004 #define DEFAULT_SEEKS 2 /* A good number if you don't know better. */ 1005 extern void register_shrinker(struct shrinker *); 1006 extern void unregister_shrinker(struct shrinker *); 1007 1008 int vma_wants_writenotify(struct vm_area_struct *vma); 1009 1010 extern pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr, spinlock_t **ptl); 1011 1012 #ifdef __PAGETABLE_PUD_FOLDED 1013 static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, 1014 unsigned long address) 1015 { 1016 return 0; 1017 } 1018 #else 1019 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address); 1020 #endif 1021 1022 #ifdef __PAGETABLE_PMD_FOLDED 1023 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud, 1024 unsigned long address) 1025 { 1026 return 0; 1027 } 1028 #else 1029 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address); 1030 #endif 1031 1032 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address); 1033 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address); 1034 1035 /* 1036 * The following ifdef needed to get the 4level-fixup.h header to work. 1037 * Remove it when 4level-fixup.h has been removed. 1038 */ 1039 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK) 1040 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address) 1041 { 1042 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))? 1043 NULL: pud_offset(pgd, address); 1044 } 1045 1046 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address) 1047 { 1048 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))? 1049 NULL: pmd_offset(pud, address); 1050 } 1051 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */ 1052 1053 #if USE_SPLIT_PTLOCKS 1054 /* 1055 * We tuck a spinlock to guard each pagetable page into its struct page, 1056 * at page->private, with BUILD_BUG_ON to make sure that this will not 1057 * overflow into the next struct page (as it might with DEBUG_SPINLOCK). 1058 * When freeing, reset page->mapping so free_pages_check won't complain. 1059 */ 1060 #define __pte_lockptr(page) &((page)->ptl) 1061 #define pte_lock_init(_page) do { \ 1062 spin_lock_init(__pte_lockptr(_page)); \ 1063 } while (0) 1064 #define pte_lock_deinit(page) ((page)->mapping = NULL) 1065 #define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));}) 1066 #else /* !USE_SPLIT_PTLOCKS */ 1067 /* 1068 * We use mm->page_table_lock to guard all pagetable pages of the mm. 1069 */ 1070 #define pte_lock_init(page) do {} while (0) 1071 #define pte_lock_deinit(page) do {} while (0) 1072 #define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;}) 1073 #endif /* USE_SPLIT_PTLOCKS */ 1074 1075 static inline void pgtable_page_ctor(struct page *page) 1076 { 1077 pte_lock_init(page); 1078 inc_zone_page_state(page, NR_PAGETABLE); 1079 } 1080 1081 static inline void pgtable_page_dtor(struct page *page) 1082 { 1083 pte_lock_deinit(page); 1084 dec_zone_page_state(page, NR_PAGETABLE); 1085 } 1086 1087 #define pte_offset_map_lock(mm, pmd, address, ptlp) \ 1088 ({ \ 1089 spinlock_t *__ptl = pte_lockptr(mm, pmd); \ 1090 pte_t *__pte = pte_offset_map(pmd, address); \ 1091 *(ptlp) = __ptl; \ 1092 spin_lock(__ptl); \ 1093 __pte; \ 1094 }) 1095 1096 #define pte_unmap_unlock(pte, ptl) do { \ 1097 spin_unlock(ptl); \ 1098 pte_unmap(pte); \ 1099 } while (0) 1100 1101 #define pte_alloc_map(mm, pmd, address) \ 1102 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \ 1103 NULL: pte_offset_map(pmd, address)) 1104 1105 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \ 1106 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \ 1107 NULL: pte_offset_map_lock(mm, pmd, address, ptlp)) 1108 1109 #define pte_alloc_kernel(pmd, address) \ 1110 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc_kernel(pmd, address))? \ 1111 NULL: pte_offset_kernel(pmd, address)) 1112 1113 extern void free_area_init(unsigned long * zones_size); 1114 extern void free_area_init_node(int nid, unsigned long * zones_size, 1115 unsigned long zone_start_pfn, unsigned long *zholes_size); 1116 #ifdef CONFIG_ARCH_POPULATES_NODE_MAP 1117 /* 1118 * With CONFIG_ARCH_POPULATES_NODE_MAP set, an architecture may initialise its 1119 * zones, allocate the backing mem_map and account for memory holes in a more 1120 * architecture independent manner. This is a substitute for creating the 1121 * zone_sizes[] and zholes_size[] arrays and passing them to 1122 * free_area_init_node() 1123 * 1124 * An architecture is expected to register range of page frames backed by 1125 * physical memory with add_active_range() before calling 1126 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic 1127 * usage, an architecture is expected to do something like 1128 * 1129 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn, 1130 * max_highmem_pfn}; 1131 * for_each_valid_physical_page_range() 1132 * add_active_range(node_id, start_pfn, end_pfn) 1133 * free_area_init_nodes(max_zone_pfns); 1134 * 1135 * If the architecture guarantees that there are no holes in the ranges 1136 * registered with add_active_range(), free_bootmem_active_regions() 1137 * will call free_bootmem_node() for each registered physical page range. 1138 * Similarly sparse_memory_present_with_active_regions() calls 1139 * memory_present() for each range when SPARSEMEM is enabled. 1140 * 1141 * See mm/page_alloc.c for more information on each function exposed by 1142 * CONFIG_ARCH_POPULATES_NODE_MAP 1143 */ 1144 extern void free_area_init_nodes(unsigned long *max_zone_pfn); 1145 extern void add_active_range(unsigned int nid, unsigned long start_pfn, 1146 unsigned long end_pfn); 1147 extern void remove_active_range(unsigned int nid, unsigned long start_pfn, 1148 unsigned long end_pfn); 1149 extern void remove_all_active_ranges(void); 1150 void sort_node_map(void); 1151 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn, 1152 unsigned long end_pfn); 1153 extern unsigned long absent_pages_in_range(unsigned long start_pfn, 1154 unsigned long end_pfn); 1155 extern void get_pfn_range_for_nid(unsigned int nid, 1156 unsigned long *start_pfn, unsigned long *end_pfn); 1157 extern unsigned long find_min_pfn_with_active_regions(void); 1158 extern void free_bootmem_with_active_regions(int nid, 1159 unsigned long max_low_pfn); 1160 int add_from_early_node_map(struct range *range, int az, 1161 int nr_range, int nid); 1162 void *__alloc_memory_core_early(int nodeid, u64 size, u64 align, 1163 u64 goal, u64 limit); 1164 typedef int (*work_fn_t)(unsigned long, unsigned long, void *); 1165 extern void work_with_active_regions(int nid, work_fn_t work_fn, void *data); 1166 extern void sparse_memory_present_with_active_regions(int nid); 1167 #endif /* CONFIG_ARCH_POPULATES_NODE_MAP */ 1168 1169 #if !defined(CONFIG_ARCH_POPULATES_NODE_MAP) && \ 1170 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID) 1171 static inline int __early_pfn_to_nid(unsigned long pfn) 1172 { 1173 return 0; 1174 } 1175 #else 1176 /* please see mm/page_alloc.c */ 1177 extern int __meminit early_pfn_to_nid(unsigned long pfn); 1178 #ifdef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID 1179 /* there is a per-arch backend function. */ 1180 extern int __meminit __early_pfn_to_nid(unsigned long pfn); 1181 #endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */ 1182 #endif 1183 1184 extern void set_dma_reserve(unsigned long new_dma_reserve); 1185 extern void memmap_init_zone(unsigned long, int, unsigned long, 1186 unsigned long, enum memmap_context); 1187 extern void setup_per_zone_wmarks(void); 1188 extern void calculate_zone_inactive_ratio(struct zone *zone); 1189 extern void mem_init(void); 1190 extern void __init mmap_init(void); 1191 extern void show_mem(void); 1192 extern void si_meminfo(struct sysinfo * val); 1193 extern void si_meminfo_node(struct sysinfo *val, int nid); 1194 extern int after_bootmem; 1195 1196 extern void setup_per_cpu_pageset(void); 1197 1198 extern void zone_pcp_update(struct zone *zone); 1199 1200 /* nommu.c */ 1201 extern atomic_long_t mmap_pages_allocated; 1202 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t); 1203 1204 /* prio_tree.c */ 1205 void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old); 1206 void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *); 1207 void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *); 1208 struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma, 1209 struct prio_tree_iter *iter); 1210 1211 #define vma_prio_tree_foreach(vma, iter, root, begin, end) \ 1212 for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \ 1213 (vma = vma_prio_tree_next(vma, iter)); ) 1214 1215 static inline void vma_nonlinear_insert(struct vm_area_struct *vma, 1216 struct list_head *list) 1217 { 1218 vma->shared.vm_set.parent = NULL; 1219 list_add_tail(&vma->shared.vm_set.list, list); 1220 } 1221 1222 /* mmap.c */ 1223 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin); 1224 extern int vma_adjust(struct vm_area_struct *vma, unsigned long start, 1225 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert); 1226 extern struct vm_area_struct *vma_merge(struct mm_struct *, 1227 struct vm_area_struct *prev, unsigned long addr, unsigned long end, 1228 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t, 1229 struct mempolicy *); 1230 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *); 1231 extern int split_vma(struct mm_struct *, 1232 struct vm_area_struct *, unsigned long addr, int new_below); 1233 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *); 1234 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *, 1235 struct rb_node **, struct rb_node *); 1236 extern void unlink_file_vma(struct vm_area_struct *); 1237 extern struct vm_area_struct *copy_vma(struct vm_area_struct **, 1238 unsigned long addr, unsigned long len, pgoff_t pgoff); 1239 extern void exit_mmap(struct mm_struct *); 1240 1241 extern int mm_take_all_locks(struct mm_struct *mm); 1242 extern void mm_drop_all_locks(struct mm_struct *mm); 1243 1244 #ifdef CONFIG_PROC_FS 1245 /* From fs/proc/base.c. callers must _not_ hold the mm's exe_file_lock */ 1246 extern void added_exe_file_vma(struct mm_struct *mm); 1247 extern void removed_exe_file_vma(struct mm_struct *mm); 1248 #else 1249 static inline void added_exe_file_vma(struct mm_struct *mm) 1250 {} 1251 1252 static inline void removed_exe_file_vma(struct mm_struct *mm) 1253 {} 1254 #endif /* CONFIG_PROC_FS */ 1255 1256 extern int may_expand_vm(struct mm_struct *mm, unsigned long npages); 1257 extern int install_special_mapping(struct mm_struct *mm, 1258 unsigned long addr, unsigned long len, 1259 unsigned long flags, struct page **pages); 1260 1261 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long); 1262 1263 extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr, 1264 unsigned long len, unsigned long prot, 1265 unsigned long flag, unsigned long pgoff); 1266 extern unsigned long mmap_region(struct file *file, unsigned long addr, 1267 unsigned long len, unsigned long flags, 1268 unsigned int vm_flags, unsigned long pgoff); 1269 1270 static inline unsigned long do_mmap(struct file *file, unsigned long addr, 1271 unsigned long len, unsigned long prot, 1272 unsigned long flag, unsigned long offset) 1273 { 1274 unsigned long ret = -EINVAL; 1275 if ((offset + PAGE_ALIGN(len)) < offset) 1276 goto out; 1277 if (!(offset & ~PAGE_MASK)) 1278 ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT); 1279 out: 1280 return ret; 1281 } 1282 1283 extern int do_munmap(struct mm_struct *, unsigned long, size_t); 1284 1285 extern unsigned long do_brk(unsigned long, unsigned long); 1286 1287 /* filemap.c */ 1288 extern unsigned long page_unuse(struct page *); 1289 extern void truncate_inode_pages(struct address_space *, loff_t); 1290 extern void truncate_inode_pages_range(struct address_space *, 1291 loff_t lstart, loff_t lend); 1292 1293 /* generic vm_area_ops exported for stackable file systems */ 1294 extern int filemap_fault(struct vm_area_struct *, struct vm_fault *); 1295 1296 /* mm/page-writeback.c */ 1297 int write_one_page(struct page *page, int wait); 1298 void task_dirty_inc(struct task_struct *tsk); 1299 1300 /* readahead.c */ 1301 #define VM_MAX_READAHEAD 128 /* kbytes */ 1302 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */ 1303 1304 int force_page_cache_readahead(struct address_space *mapping, struct file *filp, 1305 pgoff_t offset, unsigned long nr_to_read); 1306 1307 void page_cache_sync_readahead(struct address_space *mapping, 1308 struct file_ra_state *ra, 1309 struct file *filp, 1310 pgoff_t offset, 1311 unsigned long size); 1312 1313 void page_cache_async_readahead(struct address_space *mapping, 1314 struct file_ra_state *ra, 1315 struct file *filp, 1316 struct page *pg, 1317 pgoff_t offset, 1318 unsigned long size); 1319 1320 unsigned long max_sane_readahead(unsigned long nr); 1321 unsigned long ra_submit(struct file_ra_state *ra, 1322 struct address_space *mapping, 1323 struct file *filp); 1324 1325 /* Do stack extension */ 1326 extern int expand_stack(struct vm_area_struct *vma, unsigned long address); 1327 #ifdef CONFIG_IA64 1328 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address); 1329 #endif 1330 extern int expand_stack_downwards(struct vm_area_struct *vma, 1331 unsigned long address); 1332 1333 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */ 1334 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr); 1335 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr, 1336 struct vm_area_struct **pprev); 1337 1338 /* Look up the first VMA which intersects the interval start_addr..end_addr-1, 1339 NULL if none. Assume start_addr < end_addr. */ 1340 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr) 1341 { 1342 struct vm_area_struct * vma = find_vma(mm,start_addr); 1343 1344 if (vma && end_addr <= vma->vm_start) 1345 vma = NULL; 1346 return vma; 1347 } 1348 1349 static inline unsigned long vma_pages(struct vm_area_struct *vma) 1350 { 1351 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT; 1352 } 1353 1354 pgprot_t vm_get_page_prot(unsigned long vm_flags); 1355 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr); 1356 int remap_pfn_range(struct vm_area_struct *, unsigned long addr, 1357 unsigned long pfn, unsigned long size, pgprot_t); 1358 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *); 1359 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr, 1360 unsigned long pfn); 1361 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr, 1362 unsigned long pfn); 1363 1364 struct page *follow_page(struct vm_area_struct *, unsigned long address, 1365 unsigned int foll_flags); 1366 #define FOLL_WRITE 0x01 /* check pte is writable */ 1367 #define FOLL_TOUCH 0x02 /* mark page accessed */ 1368 #define FOLL_GET 0x04 /* do get_page on page */ 1369 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */ 1370 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */ 1371 1372 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr, 1373 void *data); 1374 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address, 1375 unsigned long size, pte_fn_t fn, void *data); 1376 1377 #ifdef CONFIG_PROC_FS 1378 void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long); 1379 #else 1380 static inline void vm_stat_account(struct mm_struct *mm, 1381 unsigned long flags, struct file *file, long pages) 1382 { 1383 } 1384 #endif /* CONFIG_PROC_FS */ 1385 1386 #ifdef CONFIG_DEBUG_PAGEALLOC 1387 extern int debug_pagealloc_enabled; 1388 1389 extern void kernel_map_pages(struct page *page, int numpages, int enable); 1390 1391 static inline void enable_debug_pagealloc(void) 1392 { 1393 debug_pagealloc_enabled = 1; 1394 } 1395 #ifdef CONFIG_HIBERNATION 1396 extern bool kernel_page_present(struct page *page); 1397 #endif /* CONFIG_HIBERNATION */ 1398 #else 1399 static inline void 1400 kernel_map_pages(struct page *page, int numpages, int enable) {} 1401 static inline void enable_debug_pagealloc(void) 1402 { 1403 } 1404 #ifdef CONFIG_HIBERNATION 1405 static inline bool kernel_page_present(struct page *page) { return true; } 1406 #endif /* CONFIG_HIBERNATION */ 1407 #endif 1408 1409 extern struct vm_area_struct *get_gate_vma(struct task_struct *tsk); 1410 #ifdef __HAVE_ARCH_GATE_AREA 1411 int in_gate_area_no_task(unsigned long addr); 1412 int in_gate_area(struct task_struct *task, unsigned long addr); 1413 #else 1414 int in_gate_area_no_task(unsigned long addr); 1415 #define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);}) 1416 #endif /* __HAVE_ARCH_GATE_AREA */ 1417 1418 int drop_caches_sysctl_handler(struct ctl_table *, int, 1419 void __user *, size_t *, loff_t *); 1420 unsigned long shrink_slab(unsigned long scanned, gfp_t gfp_mask, 1421 unsigned long lru_pages); 1422 1423 #ifndef CONFIG_MMU 1424 #define randomize_va_space 0 1425 #else 1426 extern int randomize_va_space; 1427 #endif 1428 1429 const char * arch_vma_name(struct vm_area_struct *vma); 1430 void print_vma_addr(char *prefix, unsigned long rip); 1431 1432 void sparse_mem_maps_populate_node(struct page **map_map, 1433 unsigned long pnum_begin, 1434 unsigned long pnum_end, 1435 unsigned long map_count, 1436 int nodeid); 1437 1438 struct page *sparse_mem_map_populate(unsigned long pnum, int nid); 1439 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node); 1440 pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node); 1441 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node); 1442 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node); 1443 void *vmemmap_alloc_block(unsigned long size, int node); 1444 void *vmemmap_alloc_block_buf(unsigned long size, int node); 1445 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long); 1446 int vmemmap_populate_basepages(struct page *start_page, 1447 unsigned long pages, int node); 1448 int vmemmap_populate(struct page *start_page, unsigned long pages, int node); 1449 void vmemmap_populate_print_last(void); 1450 1451 1452 enum mf_flags { 1453 MF_COUNT_INCREASED = 1 << 0, 1454 }; 1455 extern void memory_failure(unsigned long pfn, int trapno); 1456 extern int __memory_failure(unsigned long pfn, int trapno, int flags); 1457 extern int unpoison_memory(unsigned long pfn); 1458 extern int sysctl_memory_failure_early_kill; 1459 extern int sysctl_memory_failure_recovery; 1460 extern void shake_page(struct page *p, int access); 1461 extern atomic_long_t mce_bad_pages; 1462 extern int soft_offline_page(struct page *page, int flags); 1463 1464 extern void dump_page(struct page *page); 1465 1466 #endif /* __KERNEL__ */ 1467 #endif /* _LINUX_MM_H */ 1468